civeng

Great write-up. Thank you! I’m contemplating a similar RAG architecture for my engineering firm, but we’re dealing with roughly 20x the data volume (estimating around 9TB of project files, specs, and PDFs). I've been reading about Google's new STATIC framework (sparse matrix constrained decoding) and am really curious about the shift toward generative retrieval for massive speedups well beyond this approach. For those who have scaled RAG into the multi-terabyte range: is it actually worth exploring generative retrieval approaches like STATIC to bypass standard dense vector search, or is a traditional sharded vector DB (Milvus, Pinecone, etc.) still the most practical path at this scale?

I would guess the ingestion pain is still the same.

This new world is astounding.

Asking this question leads me to suspect you have no true curiosity or willingness to learn a new perspective as it's literally a Google away. And then you will reply in another direction that justifies your narrow world view that is of course justified 100% by your experience. This is not effective discourse and I see you doing it all over.

It’s a well-documented economic concept. You can find plenty on it if you're actually curious about the perspective. And understanding it thoroughly is a strong prerequisite to seriously engaging with other people with intent to learn. It's work you need do yourself.

I appreciate the acknowledgement ofthe 'No True Scotsman' trap. It is easy to define a side by its ideals (e.g., 'The Left is for the poor'), but the reality is that both sides muck it up the moment they take control.

Neither side actually supports the poor because both are funded by and literally are the wealthy masters. The evidence is in the trends/facts that for almost 50 years the wealth gap has only widened, regardless of who is in charge. At some point, we have to accept that the 'which side is right' argument is false.

I understand your point that a typical US citizen has options where they work.

My point is when all options include wage slavery it's not an actual option. That's it, a false dichotomy.

And that is what the OP is about. It's exploring a fundamentally different system which I understand is scary.

This is part of the false dichotomy you are not understanding. It's not actually a valid choice if there are no options. And that is what this is actually about, understanding how we have no good options, and what to do about it. I recommend start digging in and we all question or own understanding and acceptance of the system. Is it actually working with even a simple majority in the best position we can be?

Currently employing 130 wage slaves and unduly profiting from their margins, and not satisfied with the overall system at all.

The battle between the modern 'Left' and 'Right' is pure theater. It’s a false choice designed to keep us arguing over who holds the leash while we remain wage slaves. It would be interesting if some of those individuals could have been more convincing 150 years ago

Which brand did you go to?

Agreed. You got a bad batch perhaps.

When CA first required high efficacy lighting in kitchens via Title 24 in 2010, it was a train wreck, there were no good options and it got a bad name. Slowly they were required throughout new construction in tightening requirements every three years. Then technology caught up and the quality, flexibility, color and lumen options have far exceeded incandescent bulbs. Now people are doing things never imagined (as discussed somewhat awkwardly in the article) with LEDs spurred in part by legislation that pushed the technology from behind. Perhaps one of the few times technology didn't create the product and market. Legislation has created the market, and technology responded. Doesn't usually work out that way...

Now however, the new control requirements are ridiculous and add far more cost than will ever be recouped. Most likely since these requirements were planned back when California used 90 percent more electricity for lighting, it seems rather useless especially in the face of our unaffordable housing. Hopefully I am wrong there but it seems we went a bit too far.

I am squarely in the business and there are successful and failed startups everywhere. It is dynamic, challenging but there is defintely no shortage of people succeeding. Perhaps the startups just dont have the buzz or respect?

Many successful ones are already mentioned, like Procore for example. There are many more focusing in the data collection and modeling space including BIM. For example, there are still tremendous opportunities solving little problems with data collection and selling to Trimble. Or automating tasks in software and selling to Autodesk. Or add CAD like tools to a PDF reader and sell that for a hundred million (BlueBeam). The list is pretty endless, no shortage there. I believe the shortage is perhaps the humble humans that understand construction thoroughly AND understand technology to a level that would allows a bridge between the two to be constructed.

My favorite failed unicorn startup in this space that I believe demonstrates the difficulty of construction is Katerra. I personally dealt with them as a vendor and immediately knew they would fail even before Softbank gave them a billion dollars. The hubris was off the charts. A bunch of tech people that tried to hire people in the business but just couldn't understand the scale of difficulties or challenges. I must admit they generated a fair amount of schadenfreude in me.